ecell-devel Mailing List for E-Cell System

Hi all,

I have tentatively made the release criteria of version 3.1.100
as follows.   Give me a comment if you have any concerns/requests.
It is best to release this version around the midst of 
December.  

Bugs:
- 850579 dmcompile should set .DLL suffix to DMs on Windows.  satyanandavel
- 850115 Error compiling ecell-3.1.99 or ecell-cvs satyanandavel
- 723808 Stepper priority* satyanandavel

Feature requests:
- 748987 upgrade to empy-3.1* tsakurada
- 803053 ExpressionFluxProcess* tishida
- 836655 SBML importer tishida


-sha

Modeleditor has been merged with the main ecell3 CVS tree. From now on
if you install and build ecell3 from CVS you will also get modeleditor.
Modeleditor can be invoked by:
ecell-3-me
or
ecell3-me -f [eml filename]

commands.
As modeleditor is still in development phase it needs a lot of testing.
Please post your bug reports and feature requests on sourceforgenet
tracker ( assign it to bgabor) or post it on ecell3-devel.
All feedbacks are welcome.
Gabor

Hi,


I have renamed Process::process() with Process:fire().

You need to rename that method in your DM cpp files accordingly.
Sorry for this last minute change, but this is just a better name,
because the old name (process()) was somewhat meaningless.


-sha

Hi,


I've added a description of the bug and instructions of how to
build ecell3 on Fedora here:

http://ecell.sourceforge.net/download.html

Also, the as_to_python_function.hpp file is placed here:

http://ecell.sourceforge.net/additional-packages/


I think ecell3 itself is ready for Fedora Core 1 now.   
If the RPM of boost that fixes this problem is available when
we release the next version (3.1.100), we would consider building
the binary RPM on Fedora.   This would break the support for RedHat 9
unless we statically link some libraries such as libstdc++.so.


-sha



> Hi again,
> 
> To compile ecell-cvs or ecell-3.1.99 in Fedora, we need to replace 
> /usr/include/boost/python/converter/as_to_python_function.hpp with the 
> current cvs version of the file. This is because gcc-3.3.2 which is 
> shipped with Fedora has a bug which appears when we use the header file. 
> The cvs-version of the file has a workaround for it.

Hi ecell-devel,


I would like to name some most important developments currently ongoing 
in ecell3 project.

- SBML importer   (Ishida/zak)
- Implicit higher order ODE/DAE solvers, and an adaptive
  explicit/implicit ODE solver (kaizu)
- PythonProcess/ExpressionProcess for discrete time/event steppers
   (ishida/shafi)


Although not included in the release criteria of version 3.2,
many users are requesting these functionalities.   E-Cell 3 wouldn't
reach a really usable state without these.
I would like to see working versions of the modules by next Feb., or
possibly earlier.


After these developments, I would like to put emphases on the following
modules:

- The model editor with pathway view and SBML layouting compatibility (gabor)

- Tau-leap, Langevin, and dynamic switching of these algorithms (cheemeng)

- a SBML exporter (?? zak)

- the SMP parallel scheduler (satya)


I would expect to have some of those new functions possibly in the first half of
the next year.




-- 
Kouichi Takahashi,
Laboratory for Bioinformatics,
Institute for Advanced Biosciences,
Keio University, Japan.
http://www.sfc.keio.ac.jp/~shafi

Ishida-san,


As for the implementation of arithmetic operations of the
stack virtual machine of ExpressionFluxProcess,  we discussed
between these two ways of coding:

(1)
void ExpressionProcessBase::ADD::execute( StackMachine& aStackMachine )
  {
    *( aStackMachine.getStackPtr()-1 ) += *( aStackMachine.getStackPtr() );
    aStackMachine.getStackPtr()--;
  }

(2)
void ExpressionProcessBase::ADD::execute( StackMachine& aStackMachine )
  {
    const Real aValue( *aStackMachine.getStackPtr() );
    --aStackMachine.getStackPtr();
    *aStackMachine.getStackPtr() += aValue;
  }

gcc-3.3 yields the following assembly for x86 without SSE2:
(objdump -d ExpressionFluxProcess.so | c++filt)

(1)
0003b2c8 <libecs::ExpressionProcessBase::ADD::execute(libecs::ExpressionProcessBase::StackMachine&)>:
   3b2c8:       55                      push   %ebp
   3b2c9:       89 e5                   mov    %esp,%ebp
   3b2cb:       8b 55 0c                mov    0xc(%ebp),%edx
   3b2ce:       8b 42 0c                mov    0xc(%edx),%eax
   3b2d1:       dd 00                   fldl   (%eax)
   3b2d3:       dc 40 f8                faddl  0xfffffff8(%eax)
   3b2d6:       dd 58 f8                fstpl  0xfffffff8(%eax)
   3b2d9:       83 e8 08                sub    $0x8,%eax
   3b2dc:       89 42 0c                mov    %eax,0xc(%edx)
   3b2df:       c9                      leave
   3b2e0:       c3                      ret


(2)
0003b2c8 <libecs::ExpressionProcessBase::ADD::execute(libecs::ExpressionProcessBase::StackMachine&)>:
   3b2c8:       55                      push   %ebp
   3b2c9:       89 e5                   mov    %esp,%ebp
   3b2cb:       8b 4d 0c                mov    0xc(%ebp),%ecx
   3b2ce:       8b 41 0c                mov    0xc(%ecx),%eax
   3b2d1:       dd 00                   fldl   (%eax)
   3b2d3:       dc 40 f8                faddl  0xfffffff8(%eax)
   3b2d6:       8d 50 f8                lea    0xfffffff8(%eax),%edx
   3b2d9:       89 51 0c                mov    %edx,0xc(%ecx)
   3b2dc:       dd 58 f8                fstpl  0xfffffff8(%eax)
   3b2df:       c9                      leave
   3b2e0:       c3                      ret



Apparently the difference is subtle, except for these small things:
(1) makes use of less number of registers.   (2) ends with a 
FP opcode (fstpl), and there are two integer operations between faddl and fstpl.

I think we could just take whichever more readable.  (a matter of 
preference,, actually I like (2)).


Another (perhaps more important) thing I noticed is the cost of
invocation of these functions.  Eleven opcodes for such a simple
arithmetic operation, plus there must be some more codes necessary to
call these virtual functions.   

One way to eliminate this overhead could be using switch() in the
caller function (stackmachine's execute()) for these simple operations.

switch( )
{
   case PUSH:
     do push here;
   case ADD:
     do addition here;
   case SUB:
     ...

     ...
   default:
     call Instruction::execute();

}

This introduces additional overhead to other operations (such as CALL_FUNC),
but instead lightens these most frequently called arithmetic and push
operations.



You don't need to do this immediately;  SBML importer thing you are 
working on right now is more imminent.   
Just an idea for a possible optimization, for your next free time.


-- 
Kouichi Takahashi,
Laboratory for Bioinformatics,
Institute for Advanced Biosciences,
Keio University, Japan.
http://www.sfc.keio.ac.jp/~shafi

Hi again,

To compile ecell-cvs or ecell-3.1.99 in Fedora, we need to replace 
/usr/include/boost/python/converter/as_to_python_function.hpp with the 
current cvs version of the file. This is because gcc-3.3.2 which is 
shipped with Fedora has a bug which appears when we use the header file. 
The cvs-version of the file has a workaround for it.


-- 
Satya Nanda Vel Arjunan
Institute for Advanced Biosciences
Keio University
http://satya.host.sk/

Hi everyone,

I have made some changes to two configure scripts to enable ecell-cvs to 
be compiled in Fedora Core 1. The changes were done to accommodate 
changes in the way we obtain shared library extension from libtool 1.5 
as compared to the previous 1.4.x versions.

satya


-- 
Satya Nanda Vel Arjunan
Institute for Advanced Biosciences
Keio University
http://satya.host.sk/

Ok, I think I have fixed it.
Could you retry installing it?
Gabor
On Sat, 2003-11-15 at 17:20, Satya Arjunan wrote:
> Hi Gabor,
> 
> I have tried installing modeleditor with ./configure 
> --prefix=/home/satya, but it says that it couldn't find opengui.py when 
> I run ecell3-me. It is the same situation if I install with just 
> ./configure. I am using 3.1.99. A check of ecell3-me shows that it 
> executes "ecell3-python opengui.py". Should it also include the 
> modeleditor path?
> 
> satya
> 
> Gabor Bereczki wrote:
> 
> >Hi,
> >the first version of ModelEditor is ready. It is without pathway editor,
> >so it is still the same list style "debugging mode" editor it was at the
> >end of the summer. However redo and undo operations are now supported.
> >this is a quick guide on how to install and run it.
> >
> ># check out from CVS
> >$ cvs -d
> >:pserver:ga...@cs...:/home/ecell/wg/cvsroot
> >login
> >$ cvs -d
> >:pserver:ga...@cs...:/home/ecell/wg/cvsroot co
> >modeleditor
> >
> >#install
> >$ cd modeleditor
> >$ ./configure
> >$ make install
> >
> >#run
> >$ ecell3-me
> >
> >I'd welcome all comments. First of all whether you succeed in installing
> >it.
> >thanks,
> >Gabor
> >
> >
> >
> >-------------------------------------------------------
> >This SF.Net email sponsored by: ApacheCon 2003,
> >16-19 November in Las Vegas. Learn firsthand the latest
> >developments in Apache, PHP, Perl, XML, Java, MySQL,
> >WebDAV, and more! http://www.apachecon.com/
> >_______________________________________________
> >Ecell-devel mailing list
> >Ece...@li...
> >https://lists.sourceforge.net/lists/listinfo/ecell-devel
> >
> >
> >  
> >
> 

Hi Tom,

Unfortunately, the present release cannot load SBML model. But currently we
are developing the SBML importer which support SBML Lv. 2 and the first
implementation will be out at the end of this month (expected). 

If it is prepared, I send some information to this ecell-devel ML.

Regards,

-zak


________________________________________
From: ece...@li...
[mailto:ece...@li...] On Behalf Of
th...@co...
Sent: Saturday, November 15, 2003 7:33 AM
To: ece...@li...
Subject: [Ecell-devel] can I import SBML models into E-cell

Hi,

Can I import sbml models into E-cell? Thank you!!

Sincerely,

Tom
ece...@li...

Hi Gabor,

I have tried installing modeleditor with ./configure 
--prefix=/home/satya, but it says that it couldn't find opengui.py when 
I run ecell3-me. It is the same situation if I install with just 
./configure. I am using 3.1.99. A check of ecell3-me shows that it 
executes "ecell3-python opengui.py". Should it also include the 
modeleditor path?

satya

Gabor Bereczki wrote:

>Hi,
>the first version of ModelEditor is ready. It is without pathway editor,
>so it is still the same list style "debugging mode" editor it was at the
>end of the summer. However redo and undo operations are now supported.
>this is a quick guide on how to install and run it.
>
># check out from CVS
>$ cvs -d
>:pserver:ga...@cs...:/home/ecell/wg/cvsroot
>login
>$ cvs -d
>:pserver:ga...@cs...:/home/ecell/wg/cvsroot co
>modeleditor
>
>#install
>$ cd modeleditor
>$ ./configure
>$ make install
>
>#run
>$ ecell3-me
>
>I'd welcome all comments. First of all whether you succeed in installing
>it.
>thanks,
>Gabor
>
>
>
>-------------------------------------------------------
>This SF.Net email sponsored by: ApacheCon 2003,
>16-19 November in Las Vegas. Learn firsthand the latest
>developments in Apache, PHP, Perl, XML, Java, MySQL,
>WebDAV, and more! http://www.apachecon.com/
>_______________________________________________
>Ecell-devel mailing list
>Ece...@li...
>https://lists.sourceforge.net/lists/listinfo/ecell-devel
>
>
>  
>


-- 
Satya Nanda Vel Arjunan
Institute for Advanced Biosciences
Keio University
http://satya.host.sk/

Hi,
 
Can I import sbml models into E-cell? Thank you!!
 
Sincerely,
 
Tom
ece...@li...

Shafi,

I have reported the problem to both fedora-devel and boost developers 
mailing list. I found out from the boost ML archives that the next 
release, 1.31.0, is scheduled to be out on Nov 17 (or sometime next 
week). The current cvs version of boost has a workaround for the bug. 
When I tried compiling ecell-3.1.99 with it however, I received the 
following error messages:

/bin/sh ../libtool --mode=compile g++ -DHAVE_CONFIG_H -I. -I. -I..     
-g -O2  -I.. -I/root/src/ecell-3.1.99 -I../../libltdl 
-I/usr/include/python2.2 -ftemplate-depth-70 -c -o PyEcs.lo `test -f 
'PyEcs.cpp' || echo './'`PyEcs.cpp
g++ -DHAVE_CONFIG_H -I. -I. -I.. -g -O2 -I.. -I/root/src/ecell-3.1.99 
-I../../libltdl -I/usr/include/python2.2 -ftemplate-depth-70 -c 
PyEcs.cpp -MT PyEcs.lo -MD -MP -MF .deps/PyEcs.TPlo  -fPIC -DPIC -o 
PyEcs.lo
In file included from PyEcs.cpp:40:
PyEcs.hpp: In constructor `PythonCallable::PythonCallable(PyObject*)':
PyEcs.hpp:66: error: `throw_argument_error' undeclared in namespace `
  boost::python'
PyEcs.hpp: In static member function `static const libecs::Polymorph
  register_Polymorph_from_python::Polymorph_from_python(PyObject*)':
PyEcs.hpp:265: error: `throw_argument_error' undeclared in namespace `
  boost::python'


It appears that the latest boost (boost/python/error.hpp) no longer has 
throw_argument_error. I commented out the two occurrences of it in 
PyEcs.hpp and it compiled successfully. I have tested the heatshock 
model and it seems to be working fine on Fedora Core. For your 
information, the gcc-3.3 bug is still unresolved 
(http://gcc.gnu.org/bugzilla/show_bug.cgi?id=10858).

satya





>Satya,
>
>
>First report this problem to fedora-devel and boost people, and
>ask their timeline.   Then try boost-cvs and see there still is
>a problem.  I know this is a bug in gcc, but upgrading boost is
>better because forcing users to change such core components 
>like gcc is problematic.   If this still fails, try gcc 3.4.
>
>I am sorry to know the bug is still there in gcc 3.3.2.  I hoped
>it is fixed in that version.
>
>
>-sha
>
>
>
>  
>
>>The problem of booting anaconda was related to this bug: 
>>http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=109438
>>I have used the workaround using "linux allowcddma" and it worked 
>>without any problem. I installed all the packages in all 3 Fedora 
>>installation CDs. It seems the only package that is not included in 
>>Fedora 1 is the Numerical Python. After creating and installing an rpm 
>>package of Numeric-23.0 I was able to install ecell-3.1.99-2.i386.rpm 
>>and ecell-3.1.99-2.src.rpm without any problem.
>>
>>The problem came when I tried to compile the source, 
>>ecell-3.1.99.tar.gz. The configure itself went fine. But I faced the 
>>"method_call_expr" not supported by dump_expr" bug. As you know, this is 
>>a bug in gcc-3.3.2 when trying to compile PyEcs.cpp with boost-1.30.2. 
>>Both gcc-3.3.2 and boost-1.30.2 are standard packages of Fedora 1.
>>Let me know what I should do next.
>>
>>satya
>>
>>
>>
>>
>>-------------------------------------------------------
>>This SF.Net email sponsored by: ApacheCon 2003,
>>16-19 November in Las Vegas. Learn firsthand the latest
>>developments in Apache, PHP, Perl, XML, Java, MySQL,
>>WebDAV, and more! http://www.apachecon.com/
>>_______________________________________________
>>Ecell-devel mailing list
>>Ece...@li...
>>https://lists.sourceforge.net/lists/listinfo/ecell-devel
>>    
>>


-- 
Satya Nanda Vel Arjunan
Institute for Advanced Biosciences
Keio University
http://satya.host.sk/

Satya,


First report this problem to fedora-devel and boost people, and
ask their timeline.   Then try boost-cvs and see there still is
a problem.  I know this is a bug in gcc, but upgrading boost is
better because forcing users to change such core components 
like gcc is problematic.   If this still fails, try gcc 3.4.

I am sorry to know the bug is still there in gcc 3.3.2.  I hoped
it is fixed in that version.


-sha



> The problem of booting anaconda was related to this bug: 
> http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=109438
> I have used the workaround using "linux allowcddma" and it worked 
> without any problem. I installed all the packages in all 3 Fedora 
> installation CDs. It seems the only package that is not included in 
> Fedora 1 is the Numerical Python. After creating and installing an rpm 
> package of Numeric-23.0 I was able to install ecell-3.1.99-2.i386.rpm 
> and ecell-3.1.99-2.src.rpm without any problem.
> 
> The problem came when I tried to compile the source, 
> ecell-3.1.99.tar.gz. The configure itself went fine. But I faced the 
> "method_call_expr" not supported by dump_expr" bug. As you know, this is 
> a bug in gcc-3.3.2 when trying to compile PyEcs.cpp with boost-1.30.2. 
> Both gcc-3.3.2 and boost-1.30.2 are standard packages of Fedora 1.
> Let me know what I should do next.
> 
> satya
> 
> 
> 
> 
> -------------------------------------------------------
> This SF.Net email sponsored by: ApacheCon 2003,
> 16-19 November in Las Vegas. Learn firsthand the latest
> developments in Apache, PHP, Perl, XML, Java, MySQL,
> WebDAV, and more! http://www.apachecon.com/
> _______________________________________________
> Ecell-devel mailing list
> Ece...@li...
> https://lists.sourceforge.net/lists/listinfo/ecell-devel
-- 
Kouichi Takahashi,
Laboratory for Bioinformatics,
Institute for Advanced Biosciences,
Keio University, Japan.
http://www.sfc.keio.ac.jp/~shafi

Shafi,

The problem of booting anaconda was related to this bug: 
http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=109438
I have used the workaround using "linux allowcddma" and it worked 
without any problem. I installed all the packages in all 3 Fedora 
installation CDs. It seems the only package that is not included in 
Fedora 1 is the Numerical Python. After creating and installing an rpm 
package of Numeric-23.0 I was able to install ecell-3.1.99-2.i386.rpm 
and ecell-3.1.99-2.src.rpm without any problem.

The problem came when I tried to compile the source, 
ecell-3.1.99.tar.gz. The configure itself went fine. But I faced the 
"method_call_expr" not supported by dump_expr" bug. As you know, this is 
a bug in gcc-3.3.2 when trying to compile PyEcs.cpp with boost-1.30.2. 
Both gcc-3.3.2 and boost-1.30.2 are standard packages of Fedora 1.
Let me know what I should do next.

satya

Hi,
the first version of ModelEditor is ready. It is without pathway editor,
so it is still the same list style "debugging mode" editor it was at the
end of the summer. However redo and undo operations are now supported.
this is a quick guide on how to install and run it.

# check out from CVS
$ cvs -d
:pserver:ga...@cs...:/home/ecell/wg/cvsroot
login
$ cvs -d
:pserver:ga...@cs...:/home/ecell/wg/cvsroot co
modeleditor

#install
$ cd modeleditor
$ ./configure
$ make install

#run
$ ecell3-me

I'd welcome all comments. First of all whether you succeed in installing
it.
thanks,
Gabor